Databricks Unveils CustomerLake to Replace Legacy CDPs With AI Agents

Introducing CustomerLake – the agentic CDP built in Databricks — Databricks

Databricks

As autonomous consumer agents compress purchase decisions into seconds, enterprises must abandon batch campaign waterfalls for lakehouse-native AI agents executing continuous one-to-one personalization.

AI-assisted summary of Introducing CustomerLake – the agentic CDP built in Databricks — original source. Check the original for full context.

Source published:

Key ideas

  • Databricks introduced CustomerLake, an agentic Customer Data Platform (CDP) built directly within the Databricks Lakehouse and governed under Unity Catalog without extracting data into third-party silos. (8:55)
  • CustomerLake replaces traditional multi-week marketing waterfall campaigns with "Infinity Campaigns," deploying autonomous agents to deliver continuous, real-time 1:1 personalized decisioning for an audience of one. (3:51)
  • Profile Agents automate Customer 360 data preparation and feature an agentic identity resolution loop combining automated rule generation, LLM reasoning for ambiguous matches, and human-in-the-loop review. (11:43)
  • Campaign Agents leverage Databricks Genie for natural language audience segmentation and launch with native, fully managed Reverse ETL connectors supporting roughly 20 major marketing and advertising endpoints. (13:54)
  • CustomerLake is officially available in private preview, supported by an ecosystem of data, activation, and systems integration partners. (21:06)

Detailed notes

## Background Published by Databricks on YouTube on June 24, 2026, this keynote presentation introduces CustomerLake, an agentic customer data platform built directly inside Databricks. According to the video description, CustomerLake combines Customer 360, identity resolution, audience building, campaign automation, activation, and personalization into a single governed lakehouse environment, enabling marketing teams to deploy autonomous AI agents that run continuous 1:1 campaigns.

## TL;DR Tasso Argyros and Justin announce CustomerLake, an agentic CDP natively embedded within the Databricks Lakehouse and governed by Unity Catalog to overcome the latency and silo limitations of legacy waterfall marketing. The platform introduces 'Infinity Campaigns,' which replace static multi-million-person cohorts with autonomous agents delivering real-time, evergreen 1:1 personalization. CustomerLake splits operational workloads between Profile Agents—which automate data ingestion, hygiene, and agentic identity resolution—and Campaign Agents, which empower marketers to generate briefs, build natural-language audiences via Databricks Genie, and activate data using managed Reverse ETL connectors. A live demonstration models in-venue fan engagement for Brooks Sports, showing transparent agent decisioning across deferrals, match day guides, and personalized upgrade offers.

## Atmosphere & style The presentation carries the polished, high-energy cadence of a flagship enterprise tech keynote, balancing strategic industry critique with tangible product demonstrations. Argyros speaks with personal conviction and seasoned authority, framing the transition to agentic commerce not merely as an incremental software upgrade, but as the culmination of his twenty-year career bridging database engines and martech applications. The tone shifts seamlessly from visionary declarations about autonomous buyer agents to live, tactical software demos, creating an engaging blend of forward-looking ambition and practical enterprise execution.

## Bulletin - [08:55](https://www.youtube.com/watch?v=Po1gXA67wdc&t=535s) Databricks introduced CustomerLake, an agentic Customer Data Platform (CDP) built directly within the Databricks Lakehouse and governed under Unity Catalog without extracting data into third-party silos. - [03:51](https://www.youtube.com/watch?v=Po1gXA67wdc&t=231s) CustomerLake replaces traditional multi-week marketing waterfall campaigns with "Infinity Campaigns," deploying autonomous agents to deliver continuous, real-time 1:1 personalized decisioning for an audience of one. - [11:43](https://www.youtube.com/watch?v=Po1gXA67wdc&t=703s) Profile Agents automate Customer 360 data preparation and feature an agentic identity resolution loop combining automated rule generation, LLM reasoning for ambiguous matches, and human-in-the-loop review. - [13:54](https://www.youtube.com/watch?v=Po1gXA67wdc&t=834s) Campaign Agents leverage Databricks Genie for natural language audience segmentation and launch with native, fully managed Reverse ETL connectors supporting roughly 20 major marketing and advertising endpoints. - [21:06](https://www.youtube.com/watch?v=Po1gXA67wdc&t=1266s) CustomerLake is officially available in private preview, supported by an ecosystem of data, activation, and systems integration partners.

## Key takeaways - [00:01](https://www.youtube.com/watch?v=Po1gXA67wdc&t=1s) Artificial intelligence is shifting consumer commerce from conversational search queries to autonomous agent delegation, accelerating purchase decisions from weeks to minutes across headless channels. - [02:28](https://www.youtube.com/watch?v=Po1gXA67wdc&t=148s) Legacy enterprise marketing workflows structured around slow waterfall dependencies—from data engineering to creative production—are structurally unable to keep pace with machine-speed autonomous buyers. - [03:51](https://www.youtube.com/watch?v=Po1gXA67wdc&t=231s) Infinity Campaigns transition marketing from multi-million-person cohorts to dedicated autonomous agents serving an audience of one through persistent feedback loops. - [03:51](https://www.youtube.com/watch?v=Po1gXA67wdc&t=231s) In an agentic marketing framework, human practitioners are elevated rather than replaced, acting as chief strategists, orchestrators, and approvers of autonomous agent actions. - [06:41](https://www.youtube.com/watch?v=Po1gXA67wdc&t=401s) Legacy CDPs suffer from a structural disconnect as middleware detached from the core enterprise data lake, leading to high latency, security sprawl, and context deficits. - [06:41](https://www.youtube.com/watch?v=Po1gXA67wdc&t=401s) An Agentic CDP requires three architectural imperatives: native embedding within the data lakehouse, foundation models at the architectural core, and native design for continuous Infinity Campaigns. - [08:55](https://www.youtube.com/watch?v=Po1gXA67wdc&t=535s) CustomerLake is divided into two primary modules: Profile Agents to clean and assemble raw data into Customer 360 profiles, and Campaign Agents to orchestrate 1:1 marketing campaigns. - [08:55](https://www.youtube.com/watch?v=Po1gXA67wdc&t=535s) Because CustomerLake resides natively inside Databricks, customer data inherits all access control, lineage, and compliance policies defined in Unity Catalog without data movement. - [10:18](https://www.youtube.com/watch?v=Po1gXA67wdc&t=618s) Profile Agents generate end-to-end data engineering pipelines from natural language prompts, using Lakeflow Connect for SaaS ingestion and Databricks Marketplace for data hydration. - [11:43](https://www.youtube.com/watch?v=Po1gXA67wdc&t=703s) Agentic identity resolution overcomes the brittleness of static SQL rules and standard ML by using LLMs to evaluate ambiguous matches and continuously refining its rules via human feedback. - [11:43](https://www.youtube.com/watch?v=Po1gXA67wdc&t=703s) CustomerLake integrates with Databricks Clean Rooms, enabling brands to enrich and match first-party records with external partner networks without exposing raw customer data. - [13:54](https://www.youtube.com/watch?v=Po1gXA67wdc&t=834s) By building Infinity Campaigns directly on large language models, enterprises can execute 1:1 personalization immediately out of the box without spending months training bespoke ML models. - [13:54](https://www.youtube.com/watch?v=Po1gXA67wdc&t=834s) Native managed Reverse ETL connectors allow marketers to activate audiences and customer attributes across approximately 20 major marketing and ad platforms with a single click. - [16:09](https://www.youtube.com/watch?v=Po1gXA67wdc&t=969s) Natural language prompts entered into Databricks Genie automatically generate auditable, production-grade SQL queries to segment hundreds of thousands of users instantaneously. - [18:42](https://www.youtube.com/watch?v=Po1gXA67wdc&t=1122s) Infinity Campaigns provide full transparency into agentic reasoning, demonstrating explainable decisions such as intentionally deferring outreach when frequency caps or customer signals dictate inaction.

## Detailed walkthrough ### [00:01](https://www.youtube.com/watch?v=Po1gXA67wdc&t=1s) Background and the AI-Driven Shift in Customer Experience

Tasso Argyros opens the presentation by reflecting on his career trajectory, explaining why introducing this new platform feels like a return to his roots. He spent the last twenty years launching and scaling two distinct enterprises: first dedicating a decade to building a database infrastructure company, and then spending the subsequent ten years establishing a marketing technology company aimed at uniting data with an accessible business user interface. His core passion has consistently centered on placing data directly into the hands of business operators. The transition to generative AI now makes it possible to combine foundational data platforms with intelligent business agents.

Turning to the state of modern enterprise competition, Argyros notes that superior customer experience has become the decisive battleground for brands. However, customer interactions are currently undergoing a fundamental transformation because artificial intelligence is completely altering how consumers evaluate and purchase products. He polls the audience on how many people already consult a large language model before making a purchase, estimating from the show of hands that roughly 80 percent of attendees do so. Looking forward, he explains that buying will rapidly progress past simple chatbot consultations into full delegation, where autonomous consumer agents handle deep research, price comparisons, shipping logistics, and transaction execution. Because autonomous software operates via application programming interfaces across headless commerce platforms, decisions that previously took days or weeks during traditional buyer journeys will soon execute across multiple channels simultaneously in seconds or minutes. As Tasso Argyros puts it: “This means that buying decisions are now happening in seconds or minutes, not days or weeks which was a buyer journey until now.”

- **Career Background**: Ten years developing database engines followed by ten years developing marketing technology and business-facing UIs. - **Enterprise Competition**: 80 percent of leading companies report competing primarily on the quality of customer experiences. - **Agentic Buying**: Autonomous buyer agents will soon research products, negotiate pricing, verify shipping, and complete transactions simultaneously across headless commerce channels.

[02:28](https://www.youtube.com/watch?v=Po1gXA67wdc&t=148s) The Limits of Waterfall Marketing in the Age of Agentic Buying

Argyros contrasts the emerging environment of near-instantaneous agentic commerce with the slow, linear workflows that have governed enterprise marketing for decades. Under the traditional paradigm, marketing teams identify a broad universe of customers and partition them into massive demographic or behavioral cohorts containing millions of individuals. These broad audiences are subsequently targeted through rigid, sequential marketing campaigns engineered around a slow enterprise waterfall.

In this conventional waterfall structure, marketing leadership first outlines the campaign strategy. Next, data engineering teams spend weeks pulling together the necessary Customer 360 tables and unified data models. Marketing operations then steps in to define and construct the target audiences, followed by creative teams producing assets and copy. Marketers finally activate the collateral across external communication channels, monitor delivery, and calculate retrospective measurement metrics. While this operational model was functional when customer purchase cycles took weeks, Argyros stresses that its rigidity and multi-week turnaround times make it completely incompatible with machine-speed consumer agents. Emphasizing the breakdown of this legacy cadence, Tasso Argyros argues: “When agents buy on multiple channels at the same time, in seconds, you can't spend weeks designing a single campaign.”

- **Traditional Cohorts**: Built around broad segments containing millions of individuals rather than individual buyers. - **Waterfall Dependency Chain**: Strategy $\rightarrow$ Customer 360 data engineering $\rightarrow$ marketing operations audience generation $\rightarrow$ creative design $\rightarrow$ activation $\rightarrow$ reporting and measurement. - **Latency Mismatch**: Legacy waterfall campaign cycles require weeks or months to adjust, failing to match buyer decisions occurring in fractions of a minute.

[03:51](https://www.youtube.com/watch?v=Po1gXA67wdc&t=231s) Inventing Infinity Campaigns: One-to-One Agentic Marketing

To match the operational velocity of autonomous buyers, Argyros introduces an entirely new conceptual framework: shifting from batch-and-blast cohort marketing to an "audience of one" personalized continuously in real time. Rather than relying on human staff to handcraft campaigns for millions of people, companies can deploy autonomous software agents for every individual customer. In this operational model, an assigned enterprise agent ingests the customer's latest behavioral signals, evaluates the optimal next best action, and generates tailored content specifically for that single individual. As the customer reacts and transacts, those interactions emit fresh telemetry that loops directly back to the agent in a continuous feedback cycle.

```mermaid flowchart LR accTitle: The Infinity Campaign Loop accDescr: An continuous feedback loop where an individual agent evaluates customer telemetry, personalizes content, delivers it, and processes subsequent customer actions.

subgraph AgenticCore["Agent Processing"] direction TB A["Customer Signals & Telemetry"] --> B["Personalization Agent"] B --> C["Next Best Action & Dynamic Content"] end

subgraph CustomerTouchpoint["Customer Engagement"] direction TB D["Individual Customer (Audience of One)"] --> E["Transaction & Feedback"] end

C --> D E --> A ```

Databricks terms this operational model "Infinity Campaigns." Argyros explains that the name refers both to the continuous loop visual and to the evergreen, persistent nature of the campaigns. Because autonomous agents continuously adapt to evolving consumer behaviors and shifting corporate objectives, an Infinity campaign never has to shut down or start from scratch. The system dynamically generates an infinite variety of tailored content variants, delivering immediate responses without manual lag. Crucially, Argyros clarifies that human marketers are not replaced; rather, their roles are elevated to executive oversight. As Tasso Argyros emphasizes: “The marketer is the chief strategist, the chief orchestrator, and the chief approver of what the agents are doing.”

- **Audience of One**: Replacing multi-million-person cohorts with a dedicated autonomous agent serving every individual customer. - **Evergreen Execution**: Campaigns run perpetually without fixed start and end dates, constantly self-tuning as business objectives and customer data evolve. - **Infinite Variation**: Dynamic asset and copy generation tailored specifically to the individual recipient's immediate context. - **Human Governance**: Marketers set strategic boundaries, orchestrate workflows, and approve autonomous agent actions rather than manually executing tasks.

[06:41](https://www.youtube.com/watch?v=Po1gXA67wdc&t=401s) From Legacy CDPs to the Agentic CDP

Argyros explains how historical attempts to solve customer data integration led to the creation of Customer Data Platforms (CDPs). Positioned as middleware between enterprise data stores and frontline marketing execution tools, standard CDPs ingest raw records, structure them into a Customer 360 view, provide a non-technical user interface where marketers can construct cohorts and journeys, and dispatch audience segments to activation endpoints such as email platforms and digital ad networks via reverse ETL.

However, Argyros argues that legacy CDPs suffer from a fatal structural flaw: they exist as disconnected middleware detached from the primary enterprise data store. Orchestrating intelligent agents requires direct access to high-fidelity data along with rich contextual knowledge located precisely where the agents operate. To support Infinity Campaigns, businesses need an "Agentic CDP." Argyros defines three non-negotiable requirements for this emerging software category: 1. **Embedded in the Lakehouse**: It must reside natively within the enterprise data lakehouse where the underlying data, historical context, and governance policies already exist. 2. **LLMs at the Core Architecture**: Large language models cannot be treated as superficial helper copilots; they must form the foundational architectural engine driving profile assembly and campaign logic. 3. **Engineered for Infinity Campaigns**: It must natively support evergreen, real-time, self-adapting campaigns executing one-to-one personalization at machine speed.

Underscoring the distinction between superficial generative add-ons and core platform re-architecture, Tasso Argyros asserts: “This is not just about having some co-pilot to help you do some operations better.”

- **Legacy CDP Role**: Acted as middleware extracting data out of enterprise warehouses to build static segments and pushing them to endpoints via reverse ETL. - **Middleware Disconnect**: Traditional platforms decouple business logic from the source data lake, creating latency, security sprawl, and context deficits. - **Agentic CDP Standard**: Natively embedded within lakehouse storage, built ground-up on generative models, and purpose-built for continuous 1:1 campaigns.

[08:55](https://www.youtube.com/watch?v=Po1gXA67wdc&t=535s) Introducing CustomerLake: Architecture and Core Modules

Databricks is entering the newly defined agentic CDP category with CustomerLake, an agentic customer data platform engineered natively for the Databricks platform. By building CustomerLake directly on Databricks, the platform eliminates the requirement to extract customer records into third-party silos. Data remains within the lakehouse, instantly inheriting all access controls, data lineage, and audit governance rules already defined in Unity Catalog. CustomerLake is designed with a business-first user interface so non-technical marketing practitioners can configure sophisticated workflows without writing code, while the underlying architecture scales to manage billions of autonomous personalization interactions every single day. As Tasso Argyros declares: “It can power one-to-one personalization agents to provide the perfect customer experience a billion times a day.”

CustomerLake is structured around two specialized modules: - **Profile Agents**: Autonomous backend agents that transform disparate, unformatted enterprise tables into business-ready Customer 360 profiles. - **Campaign Agents**: A marketer-facing application environment where business teams define, guide, and orchestrate automated Infinity Campaigns powered by autonomous agents.

```mermaid flowchart TD accTitle: CustomerLake Architectural Overview accDescr: The high-level architecture of CustomerLake showing the underlying Databricks Lakehouse, Unity Catalog governance, and the two operational modules: Profile Agents and Campaign Agents.

subgraph DatabricksFoundation["Databricks Lakehouse Foundation"] UC["Unity Catalog Governance & Security"] LakeData["Lakehouse Native Storage (No Data Movement)"] end

subgraph CustomerLakePlatform["CustomerLake (Agentic CDP)"] direction TB PA["Profile Agents (Raw Data to Customer 360)"] CA["Campaign Agents (Infinity Campaigns UI & Execution)"] end

subgraph BusinessOutcomes["Downstream Execution"] PersAgents["1:1 Personalization Agents (1B+ Interactions/Day)"] end

DatabricksFoundation --> CustomerLakePlatform PA --> CA CA --> PersAgents ```

- **Native Lakehouse Residency**: Customer data never leaves the secure lakehouse environment, avoiding third-party data copies. - **Unified Governance**: Inherits permissions, row-level access filters, and compliance policies established inside Databricks Unity Catalog. - **Modular Division**: Profile Agents handle data ingestion, semantic modeling, and hygiene, while Campaign Agents manage marketing strategy, content generation, and execution.

[10:18](https://www.youtube.com/watch?v=Po1gXA67wdc&t=618s) Profile Agents: Automating Customer 360 and Identity Resolution

Argyros details the mechanics of the Profile Agents module, which automates the arduous data engineering work required to turn raw, dirty records into clean Customer 360 profiles. Rather than demanding manual pipeline development from data engineering teams, Profile Agents initiate data prep workflows directly from natural language prompts specifying the target marketing use case. Once instructed, the agents ingest tables, execute transformations, append enrichment data, perform identity resolution, and continuously evaluate data quality across the profile graph in an autonomous improvement loop.

To source customer information, the platform uses Lakeflow Connect to ingest third-party application feeds or natively references tables already resident within the lakehouse. Once ingested, the agents analyze schema metadata, apply semantic tags, and perform automated hygiene routines—such as standardizing telephone numbers, normalizing postal addresses, and repairing malformed email strings. Organizations can further augment internal data via "data hydration," enriching internal profiles with third-party datasets procured directly through the Databricks Marketplace.

The central capability of Profile Agents is identity resolution across first-party customer datasets that lack a unified identifier. Argyros clarifies that modern enterprises maintain customer interactions across numerous distinct platforms—point-of-sale systems, e-commerce portals, mobile apps, and customer service desks—where user names, physical locations, and email handles often contain subtle discrepancies or formatting differences. Providing a formal operational definition, Tasso Argyros explains: “Identity resolution for first-party data, meaning data you already own, this is data you have collected from your own customers, is a process of identifying the same customer across different data sets that do not share the same customer key.”

- **Prompt-Driven Pipeline Assembly**: Profile Agents generate end-to-end data pipelines from natural language prompts describing business objectives. - **Lakeflow Connect Integration**: Ingests disparate enterprise SaaS application data alongside native Databricks Lakehouse tables. - **Automated Data Hygiene**: Semantically classifies data attributes while resolving syntax errors across emails, phone numbers, and physical addresses. - **Data Hydration**: Enriches existing internal profiles using external data assets via Databricks Marketplace. - **First-Party Identity Resolution**: Accurately maps disparate customer fragments into a single golden record without requiring consistent pre-existing primary keys.

[11:43](https://www.youtube.com/watch?v=Po1gXA67wdc&t=703s) Agentic Identity Resolution and Profile Agents

Tasso Argyros explains the critical necessity of identity resolution, emphasizing that without probabilistic matching, organizations fail to recognize repeat visitors across channels: “This is very important because if you don't do this, you may end up thinking the same customer is different people, and that leads to massive problems for marketing and customer experience.” To solve this fundamental breakdown, Databricks is introducing agentic identity resolution. Rather than relying solely on brittle rules or heavy manual coding, specialized agents generate initial identity resolution rules. When edge cases arise where rules prove inconclusive, large language models intervene to evaluate the ambiguous matches. If uncertainty remains, the system smoothly falls back to human reviewers for feedback. Crucially, this creates an active agentic feedback loop: as more identities are confirmed and resolved, the underlying agents learn, sharpening the automated rules and fine-tuning the LLM's predictive capability over time.

Argyros contrasts this adaptive framework with legacy mechanisms, noting, “Historically, this has happened through static SQL rules or machine learning. Both approaches have significant limitations, and we think the agentic identity resolution wins on all fronts over these traditional approaches.” Moving beyond internal datasets, Argyros details how CustomerLake handles third-party partner records through Databricks Clean Rooms. This enables brands to reconcile customer profiles against partner enrichment networks without their raw enterprise data ever leaving the lakehouse, delivering a privacy-first integration architecture. The final phase of Profile Agents takes these ingested, standardized, cleansed, and unified identities and transforms them into domain-specific data products through schema mapping, survivorship policies, and automated validation.

```mermaid flowchart TD accTitle: Agentic Identity Resolution Loop accDescr: A decision flow showing rule evaluation falling back to LLMs, human feedback, and a continuous learning loop. subgraph Resolution["Resolution Engine"] A["Incoming Customer Records"] --> B{"Agentic Rules"} B -->|"Matched"| C["Unified Profile"] B -->|"Inconclusive"| D{"LLM Evaluation"} D -->|"Resolved"| C D -->|"Ambiguous"| E["Human Feedback"] E --> C end subgraph Feedback["Learning Engine"] E -.->|"Train & Refine"| F["Agentic Learning Loop"] D -.->|"Fine-tune Patterns"| F F -.->|"Update"| B end ```

[13:54](https://www.youtube.com/watch?v=Po1gXA67wdc&t=834s) Campaign Agents, Infinity Campaigns, and Managed Reverse ETL

Transitioning to activation, Argyros introduces Campaign Agents, which automate end-to-end campaign operations. Given a simple natural language prompt, these agents ingest business goals and broader context to generate a comprehensive campaign brief. Once human marketers review and sign off, the platform deploys agents to execute personalization at scale. Beyond cutting-edge automation, Campaign Agents still handle standard customer data platform duties like segmentation and audiencing, but modernize them through Databricks Genie. Marketers can build audiences strictly via natural language prompts or refine them inside a business UI, turning audience segments into active multi-channel journeys with a single click.

At the core of Campaign Agents lies Infinity Campaigns. Argyros underscores how leveraging foundation models removes the traditional roadblocks that have plagued enterprise personalization: “The beauty of Infinity campaigns, because it's built around the LLMs, is that you can start personalizing immediately. You don't have to spend months building ML models.” Teams no longer have to wait for bespoke machine learning decisioning engines to slowly train on historical data. Infinity Campaigns functions out of the box with built-in simulation runs, automated governance guardrails, and granular performance measurement. Furthermore, addressing a long-standing user request, Databricks announced native, fully managed Reverse ETL connectors. Launching with approximately 20 turnkey integrations across major advertising, marketing, and customer experience platforms, CustomerLake allows single-click data syncing, with dozens more connectors slated by the end of the year.

[16:09](https://www.youtube.com/watch?v=Po1gXA67wdc&t=969s) Live Demo: Audience Generation with Databricks Genie

Justin takes the stage to demo Campaign Agents and Infinity Campaigns in action, stepping into the shoes of a marketing manager for Brooks Sports. Starting from the central campaign dashboard—which tracks active campaigns, audience trends, and operational Reverse ETL syncs—Justin navigates to the audiences tab to construct a brand-new segment without touching traditional filter builder tools. Instead of assembling manual boolean conditions, he pastes a natural language query directly into the Genie interface: `"Include our most valuable Team USA fans who have spent more than $500 on merch this season, and exclude anyone who already bought the USA fan kit."`

Genie evaluates existing schema attributes and segments to dynamically generate an audience of 800,000 qualifying customers. While marketers can accept or edit the qualification criteria in the UI, Justin inspects the underlying technical execution: “You can view the SQL that is backing this audience. So, this is starting to look probably more familiar uh to many of you in the audience. Uh if you've ever supported marketing by pulling lists and writing custom campaigns, um you can get really complex SQL um to power some of these campaigns, and we got this from one prompt with Genie.” Saving the segment automatically spins up an interactive audience dashboard with Genie-generated insights, ready to immediately trigger a new Reverse ETL sync or anchor a multi-channel campaign.

[18:42](https://www.youtube.com/watch?v=Po1gXA67wdc&t=1122s) Live Demo: 1:1 Decisioning in Infinity Campaigns

Justin transitions to inspecting a live Infinity Campaign configured for an in-venue fan experience. In the campaign management console, marketers establish the high-level guardrails: audience inclusions, conversion metrics, pacing rules, and the asset library—such as a match day guide and a match day ticket upgrade offer. Justin pivots to the agents tab, which exposes how the agent evaluates individual customer profiles on a true 1:1 basis in real time.

Demonstrating the transparent reasoning behind agent decisions, Justin walks through several distinct user profiles:

- **Marcus (Deferred Delivery):** Rather than blindly sending a blast, the agent elects to defer outreach. The UI lays out the agent's explicit rationale, showing how it weighed signal timing, historical engagement data, and frequency cap guardrails to conclude Marcus should not receive a message right now. - **Customer with Match Day Offer:** Evaluating a different profile, Genie detects an opportunity to drive incremental revenue and delivers a targeted match day offer, granting a 15% discount on an in-stadium seat upgrade. - **Customer with Match Day Guide:** For another fan profile, the agent determines a commercial upsell is inappropriate and instead delivers a practical match day guide highlighting kickoff timing, parking directions, and entry protocols.

Justin emphasizes that Campaign Agents continuously evaluate and tailor these decisions for every single individual. Marketers retain full oversight through top-level analytics dashboards that track message-level open rates, click-through rates, and conversion counts across the campaign. Concluding the walkthrough, Justin remarks, “That's a quick run through of Customer Lake Campaign Agents, the world's first agentic CDP.”

[21:06](https://www.youtube.com/watch?v=Po1gXA67wdc&t=1266s) Partner Ecosystem and Private Preview Availability

Tasso Argyros returns to close the session, declaring, “The era of true one-to-one personalization is here. It's very, very, very, very exciting.” He extends gratitude to early enterprise design partners and global innovators whose data and marketing teams battle-tested the platform's agentic workflows. Argyros also recognizes the broader ecosystem of data, content, advertising, activation, and systems integration partners who are co-launching with CustomerLake to deliver end-to-end customer experience capabilities.

CustomerLake is officially open in private preview. Argyros directs interested organizations to contact their Databricks account teams to request access and encourages attendees to explore the conference's dedicated CDP breakout tracks as well as the afternoon marketing forum.

[22:36](https://www.youtube.com/watch?v=Po1gXA67wdc&t=1356s) Learning More at Data + AI Summit and Closing Thoughts

To explore the platform further, attendees are encouraged to search for CustomerLake within the official Data + AI Summit mobile app, where they can find both the dedicated marketing forum and specialized breakout sessions. Tasso Argyros explains that these sessions offer a behind-the-scenes look directly from the engineers and product leaders who built CustomerLake from the ground up, alongside presentations from early enterprise adopters sharing their first-hand experiences and initial outcomes using the agentic CDP.

Reflecting on the launch, Argyros closes the presentation on a personal note, emphasizing how meaningful it is to unite domains that have historically operated in silos. As Tasso Argyros reflects: “It's a very special moment for me to be able to bring together the worlds of business, customer experience, and data all in one place uh to build something that provides the perfect customer experience a billion times a day.”

- **Summit App Discovery:** Attendees can search "CustomerLake" in the Data + AI Summit app to register for the marketing forum and technical breakout tracks. - **Behind the Product:** Sessions feature the core engineering and product teams responsible for building CustomerLake, detailing the architecture and underlying unified data foundation. - **Early Customer Insights:** Initial enterprise design partners and early customers present real-world deployments and their early results with the platform.

## Notable quotes > [02:28](https://www.youtube.com/watch?v=Po1gXA67wdc&t=148s) When agents buy on multiple channels at the same time, in seconds, you can't spend weeks designing a single campaign. — Tasso Argyros

> [03:51](https://www.youtube.com/watch?v=Po1gXA67wdc&t=231s) The marketer is the chief strategist, the chief orchestrator, and the chief approver of what the agents are doing. — Tasso Argyros

> [13:54](https://www.youtube.com/watch?v=Po1gXA67wdc&t=834s) The beauty of Infinity campaigns, because it's built around the LLMs, is that you can start personalizing immediately. You don't have to spend months building ML models. — Tasso Argyros

## Mentioned Tasso Argyros, Justin, Databricks, CustomerLake, Unity Catalog, Databricks Genie, Lakeflow Connect, Databricks Marketplace, Databricks Clean Rooms, Data + AI Summit, Brooks Sports

## Video description Introducing CustomerLake: The agentic CDP

CustomerLake brings core CDP capablities together to help marketing and data teams turn customer data into always-on, 1:1 experiences. With CustomerLake, Customer 360, identity resolution, audience building, campaign automation, activation, and personalization are all available in one platform. And with a shared, governed data foundation, marketers can deploy AI agents that help eliminate manual campaign and keep their focus on customers. Learn more about CustomerLake: https://www.databricks.com/blog/introducing-customerlake-agentic-cdp

00:00 — The impact of AI on the customer experience 02:28 — Why legacy marketing campaigns fall short 03:51 — The shift to Infitity Campaigns 06:41 — What's a CDP? 07:35 — The need for an Agentic CDP 08:55 — Introducing CustomerLake 15:53 — Demo: CustomerLake 21:07 — CustomerLake partners

Presenters: Tasso Argyros, VP Engineering, Databricks

[Source](https://www.youtube.com/watch?v=Po1gXA67wdc)

Reading summary

Databricks introduced CustomerLake, an agentic customer data platform engineered to replace slow, cohort-based marketing cycles with continuous, one-to-one personalization embedded directly inside the lakehouse. Presenter Tasso Argyros argued that as consumers shift toward delegating transactions to autonomous AI software, enterprise marketing must pivot from multi-week waterfall campaigns to real-time agentic workflows. Machine-Speed Commerce Breaks Traditional Marketing Cadences (0:01) Argyros stated that enterprise competition increasingly hinges on customer experience, but consumer purchasing behavior is transforming under artificial intelligence. In an audience poll, he estimated that roughly 80 percent of attendees already consult large language models before purchasing. Looking forward, Argyros forecasted that buying will transition from informational queries into full delegation, with autonomous consumer agents researching products, comparing shipping logistics, and executing transactions across headless commerce APIs in seconds or minutes. Traditional enterprise marketing remains organized around slow waterfall workflows. Teams move sequentially from high-level strategy to data engineering Customer 360 models, marketing operations constructing audiences, creative asset production, channel activation, and retrospective reporting. Argyros argued that designing multi-week campaigns around broad cohorts containing millions of individuals is fundamentally incompatible with consumer agents transacting simultaneously across multiple channels at machine speed. Moving the CDP Directly into Lakehouse Architecture (6:41) Legacy Customer Data Platforms (CDPs) operated as middleware, extracting records from enterprise warehouses, assembling static segments, and pushing audiences to marketing endpoints via reverse ETL. Argyros asserted that this architecture creates data latency, context deficits, and security sprawl. To support autonomous marketing, he argued that an "Agentic CDP" must abandon disconnected middleware, reside natively within lakehouse storage, and utilize foundation models as core architecture rather than superficial helper copilots. Built natively on Databricks, CustomerLake keeps customer records in the lakehouse without generating third-party data copies, inheriting security and governance policies from Unity Catalog. The platform divides responsibilities into two modules: Profile Agents, which automate backend data transformations into Customer 360 views, and Campaign Agents, which provide a non-technical interface where marketers configure and govern automated marketing workflows. Automated Hygiene and Adaptive Identity Resolution (10:18) Profile Agents automate the engineering tasks required to clean and structure disparate records. Initiated by natural language prompts outlining business goals, Profile Agents ingest application data via Lakeflow Connect or native lakehouse tables, execute automated data hygiene on malformed emails or addresses, and enrich records using third-party datasets from the Databricks Marketplace. To integrate external partner data securely, the system links with Databricks Clean Rooms so profiles reconcile without raw enterprise data leaving the lakehouse. For internal records lacking common primary keys across points of sale, e-commerce, and service desks, CustomerLake uses agentic identity resolution. Moving beyond rigid SQL rules or brittle machine learning models, the platform deploys agents to establish baseline rules, escalates inconclusive matches to large language models, and routes persistent ambiguities to human reviewers. Decisions feed into an iterative learning loop that refines automated rules and improves pattern evaluation over time. Infinity Campaigns and Real-Time Decisioning (13:54) For campaign execution, CustomerLake introduces "Infinity Campaigns"—evergreen campaigns targeting an "audience of one" via dedicated autonomous agents. Because large language models dynamically evaluate customer signals, Argyros noted teams can personalize outreach immediately without spending months training bespoke machine learning models. Marketers use Databricks Genie to define audiences through natural language queries, which dynamically compile into complex SQL models that can trigger native Reverse ETL syncs across external advertising and customer experience platforms. A live demonstration using a simulated Brooks Sports brand illustrated how Campaign Agents evaluate individual customer profiles in real time under defined guardrails. The system deferred outreach to an individual based on frequency capping, served a seat upgrade discount to an engaged fan, and sent practical stadium transit directions to an attendee where an upsell was deemed inappropriate. Argyros stressed that human marketers remain chief strategists and approvers, shifting focus from manual campaign mechanics to executive governance. CustomerLake is currently available in private preview with approximately 20 native Reverse ETL connectors and plans for additional integrations by year-end. While Databricks positions the platform to scale to a billion personalized interactions daily, long-term enterprise adoption will depend on operational governance, platform reliability, and how effectively marketing organizations transition from legacy campaign structures to supervising autonomous software agents.

Source quotes

[2:28] Tasso Argyros: When agents buy on multiple channels at the same time, in seconds, you can't spend weeks designing a single campaign. [3:51] Tasso Argyros: The marketer is the chief strategist, the chief orchestrator, and the chief approver of what the agents are doing. [13:54] Tasso Argyros: The beauty of Infinity campaigns, because it's built around the LLMs, is that you can start personalizing immediately. You don't have to spend months building ML models.

Original transcript

[0:01] Good morning. [0:03] How's everyone doing? [0:07] Okay. [0:08] Um [0:09] it's a real uh pleasure to be here with [0:11] you today. [0:12] This is actually very, very special for [0:15] me. I'll tell you why. Um [0:18] I spent the last 20 years building two [0:20] companies. I spent 10 years building a [0:22] database company. [0:24] But then I really wanted to do something [0:27] that would bring the power of data to [0:29] the business. So I spent another decade [0:32] building a marketing technology company. [0:35] One that would combine data [0:37] but also combine a UI for the business. [0:39] That was my passion. [0:41] And it's really special to be here [0:44] because in some ways I'm back at my [0:45] roots, but what I'm about to show you is [0:48] something that truly brings the power of [0:51] data and agents to the business. [0:54] But before we talk about this, I want to [0:56] talk to you about customer experience. [0:59] Why do we talk about it? The best [1:02] companies differentiate on customer [1:04] experience. 80% [1:06] say that they compete primarily on on [1:08] the basis of the experiences they create [1:10] for their customers. [1:12] And [1:13] we live in a very interesting time [1:15] because AI is dramatically changing how [1:19] people buy and as a result how those [1:21] experiences need to be designed. [1:24] Why is that happening? [1:26] First of all, humans do not buy on their [1:29] own anymore. [1:31] How many of you do you consult with an [1:35] LLM [1:36] before you buy anything? Be honest. [1:40] Okay, I think this is like 80%. [1:43] Uh very soon is not just going to be [1:46] advice from a chatbot. It's going to be [1:48] delegating agents to do all the [1:50] research, all the price discovery, all [1:53] the buying, figure out shipping. [1:55] Everything will happen through agents. [1:57] This means that buying decisions are now [1:59] happening in seconds or minutes, [2:03] not days or weeks which was a buyer [2:06] journey until now. [2:08] And finally, because commerce is going [2:11] headless [2:12] and agents are really good with APIs, [2:16] agents can be on multiple channels [2:18] buying, researching all at the same [2:20] time. [2:23] So, this creates new challenges, right? [2:26] To how things were done before. Because [2:28] the way things were done before is very [2:30] simple. [2:31] Someone in a company, usually marketing, [2:34] would start with a universe of [2:36] customers, create segments, we call them [2:39] audiences, that would have millions of [2:41] people, [2:42] and then activate those audiences [2:45] through campaigns. [2:47] And the way these campaigns would be [2:48] designed to begin with was through this [2:51] big waterfall model, right? So, [2:53] marketing would do the strategy, [2:55] data engineering would put the customer [2:57] 360 data together, marketing ops would [3:00] create the audiences, creative would do [3:02] content, marketers would activate, [3:04] measure, right? And monitor the campaign [3:06] as they were running. [3:09] This waterfall worked, but it had some [3:12] issues, right? First of all, it takes a [3:14] really long time and is very rigid. It's [3:17] hard to change dynamically. [3:19] Um so, for all these limitations, [3:22] though, these waterfall campaigns worked [3:24] fine up to now. [3:26] But, [3:27] do they work for the era of agentic [3:31] buying? [3:32] And the answer is no, right? Of course [3:34] not. When agents buy on multiple [3:37] channels at the same time, in seconds, [3:40] you can't spend weeks designing a single [3:43] campaign. [3:44] But, what if [3:46] what if we could personalize every [3:49] interaction for every customer in real [3:52] time all the time [3:54] for their customers and their agents. [3:56] That would be a game-changer. [3:59] But to do that we need to think [4:00] differently. [4:02] We need to invent a new type of [4:04] campaign, one that doesn't rely on this [4:06] waterfall model [4:08] and one that can personalize audiences [4:10] of one instead of audiences of millions. [4:15] What if we could do this with agents, [4:17] right? So agents create the new [4:19] challenge on the buyer side, what are [4:21] also the solutions? What if we could [4:23] have an agent for every customer? [4:26] And then what this agent would do is to [4:28] process the latest signals from every [4:31] single customer, determine what's the [4:33] next best action, and create a highly [4:36] personalized content just for that [4:38] individual customer. [4:41] And from there on, that would go back to [4:43] the customer, customer would process, [4:45] transact, and then create more signals [4:47] that would then feed back to the agent [4:49] in this infinity-like loop. [4:53] We think this is the future and this is [4:56] such a new approach to how marketing can [4:59] do campaigns that it deserved a new [5:01] name. [5:03] We called it infinity campaigns. [5:07] And we call it infinity not because the [5:09] symbol looks like an infinity symbol, [5:12] but uh importantly, these are campaigns [5:15] that are evergreen. They don't need to [5:17] have a start and an end. [5:19] Because we have agents, the agents [5:21] continuously adapt, so the same campaign [5:24] can adapt to new customer behaviors, new [5:27] business goals, and it can keep running. [5:31] Also, infinity campaigns can provide [5:35] content that's highly personalized, [5:37] right? It can pick from [5:38] a multitude infinitely large number of [5:41] variations of content for what each [5:44] individual customer needs that [5:45] particular moment. [5:47] And finally, if you employ agents, they [5:50] can respond very, very fast. Right? You [5:53] You You don't have any limits in how [5:54] fast these agents can respond. [5:57] Now, it's important to note that human [6:00] remains in control. [6:02] The marketer is the chief strategist, [6:05] the chief orchestrator, and the chief [6:07] approver [6:08] of what the agents are doing. [6:11] But now they have an army of agents that [6:13] can do all the one-to-one [6:15] personalization work that humans were [6:17] just too slow to do before. [6:21] Okay. So, hopefully, I'm convincing you [6:24] that this new concept of infinity [6:26] campaign [6:27] is very cool. But the question is, can [6:29] we make it work? How do we make this [6:31] happen? [6:32] And the way this would have happened in [6:35] the past was through CDPs. [6:39] And let me talk a little bit about what [6:40] CDPs are. For those of you that don't [6:42] know, CDP stands for customer data [6:44] platform. [6:46] And it's essentially a business [6:49] application that sits between the data [6:52] and all the marketing tools, the [6:54] customer tools that actually send the [6:55] communications to the customers. So, [6:58] what do CDPs do? First, they can turn [7:00] raw data into customer 360, organized [7:03] data ready for business use. [7:05] Then, they have a business UI where [7:07] marketers can come in and build [7:08] audiences and journeys. And finally, [7:11] they send data to marketing tools that [7:14] serve content to the end customer, like [7:16] email tools or advertising platforms. [7:19] And that process is called reverse ETL [7:22] because the data goes from the CDP into [7:26] the content or the advertising platform. [7:30] The challenge here that CDPs were [7:32] middleware, so they were detached from [7:34] the data. And it's not just about data [7:36] anymore. Now, in order to make agents [7:39] work, you need data, but you also need [7:41] context, and you need data and context [7:44] to be where the agents live. [7:46] So, we think you need a new generation [7:48] of CDPs to make Infinity campaigns work, [7:52] and we call this new generation [7:54] the agentic CDP. [7:56] The agentic CDP is not just some general [7:59] concept. There's three very specific [8:01] things that a CDP needs to do in order [8:03] to be agentic. [8:05] First, it has to be embedded in the [8:07] lakehouse [8:08] because that's where the data, the [8:10] context of the agents increasingly live. [8:14] Second, it has to be built with LLMs at [8:17] the core of its architecture. [8:19] This is not just about having some [8:21] co-pilot to help you do some operations [8:24] better. [8:25] As we discussed before, the agent is the [8:27] core of how this new type of CDP works. [8:30] And finally, it has to power Infinity [8:32] campaigns. It needs to have the [8:33] capability to have evergreen, real-time, [8:37] continuously adapting campaigns that [8:39] personalize every individual customer [8:41] with individual agents. [8:44] So, agentic CDP is a new category within [8:47] there's going to be [8:49] several new players there. [8:51] However, we are one of the very first to [8:55] launch an agentic CDP, [8:58] and we call this [9:00] Customer Lake. [9:03] Thank you. [9:05] Thank you. [9:07] Customer Lake is an agentic CDP built [9:09] specifically for the Databricks [9:11] platform. [9:13] It's embedded, meaning data never has to [9:16] leave the lakehouse, and all the [9:18] governance for data and agents that [9:21] you've done with Unity Catalog [9:23] will be transferred immediately into [9:26] Customer Lake. [9:27] It's democratized. [9:29] It's built for the business users, so it [9:31] has a business first UI, [9:33] and it is autonomous. [9:35] It can power one-to-one personalization [9:37] agents to provide the perfect customer [9:39] experience a billion times a day. [9:45] Customer Lake comes with two modules. [9:47] The first is profile agents, which [9:49] automates the process of going from raw [9:51] data to customer 360. [9:54] The second is campaign agents. That's a [9:57] business first product that allows [10:00] marketers to power in 50 campaigns with [10:02] agents. [10:04] Let's double click and talk about it a [10:05] little bit more. There's a lot of [10:07] capabilities that we're introducing here [10:10] and in the remaining of this talk, we're [10:12] going to talk a little bit more about [10:13] these two products. [10:15] Uh and also show you how it works. [10:18] Okay, let's start with profile agents. [10:21] Profile agents are agents that automate [10:22] the process of creating customer 360 [10:24] data that is business ready. [10:27] So, with the prompt that defines a use [10:29] case, profile agents get to work. They [10:32] ingest, transform, enrich data, resolve [10:36] customer identities, and evaluate [10:38] profiles to generate this intelligent [10:40] loop that keeps improving quality. [10:43] The first step is to prepare customer [10:45] data for use. With Lake Flow Connect, we [10:48] can ingest third-party data [10:50] from applications or obviously use [10:53] Lakehouse data that's already there. [10:55] Our agents will semantically tag data [10:58] and also improve quality such as fix [11:00] emails, addresses, or phone numbers. [11:03] And we can also bring third-party data [11:06] in, [11:07] which we call data hydration, through [11:09] the Databricks Marketplace. [11:13] The second step is one of the key core [11:15] features of profile agents, which is [11:17] identity resolution. [11:19] What is identity resolution? Identity [11:21] resolution for first-party data, meaning [11:24] data you already own, this is data you [11:26] have collected from your own customers, [11:29] is a process of identifying the same [11:32] customer across different data sets that [11:35] do not share the same customer key. [11:38] So, you have information about the [11:39] customers. You may have names, emails, [11:41] addresses. [11:42] They don't quite match, but [11:43] probabilistically, [11:45] you can tell it's the same customer. [11:47] This is very important because if you [11:49] don't do this, [11:50] you may end up thinking the same [11:52] customer is different people, and that [11:54] leads to massive problems for marketing [11:56] and customer experience. [11:58] So, we introduce here a new approach to [12:02] identity resolution that we call the [12:03] agentic identity resolution. [12:06] We have agents that come up with rules [12:08] on how to resolve these identities. If [12:11] these rules are inconclusive, LLMs take [12:14] over and try to make a determination, [12:16] and if needed, it falls back to human [12:18] feedback. [12:19] And as more and more identities are [12:21] being identified and resolved, the [12:23] agents learn. And they enable this [12:26] agentic loop that make the rules better, [12:29] the LLM learning better, and the whole [12:31] process continuously improves. [12:34] Historically, this has happened through [12:36] static SQL rules or machine learning. [12:39] Both approaches have significant [12:41] limitations, and we think the agentic [12:43] identity resolution [12:45] wins on all fronts over these [12:47] traditional approaches. [12:53] And when it comes to third-party data, [12:55] meaning data that comes from third-party [12:58] partners, [13:00] we are introducing today a way to [13:02] resolve identities through those [13:05] partners without your data ever leaving [13:08] the lakehouse. So, this is a [13:10] privacy-first approach to third-party [13:13] identity resolution that we enable [13:16] through the Databricks clean rooms. [13:20] And the last step in profile agents, [13:23] after having ingested, transformed, [13:25] quality checked the data, [13:27] and identify the identities, is to [13:30] output the right data products for the [13:33] right use cases. So, that involves [13:34] schema mapping, [13:36] survivorship, and also data validation [13:38] so that the agents can self-improve the [13:40] quality of the final output as needed. [13:44] So, that's profile agents helping you go [13:47] from raw data to business-ready customer [13:49] 360 faster than ever before. [13:54] So, let's talk about our second product [13:56] here, campaign agents. Campaign agents [13:58] are agents that automate campaign work. [14:02] So, given a prompt, the agents can [14:03] create the right campaign uh brief [14:07] from business context and business [14:08] goals, the marketers review it and [14:10] approve it, and then we deploy agents to [14:13] do personalization. [14:17] What's cool also is that campaign agents [14:20] also do traditional CDP functionality [14:22] like audiencing. [14:24] But, of course, we do it in a way that's [14:26] Genie-powered. [14:28] So, everything can happen through [14:29] prompts, but we also have a business UI [14:32] that marketing and marketing operations [14:34] can very quickly review the work of the [14:36] agents, approve it, and you with one [14:39] click you can take audiences and turn [14:41] them into campaigns. [14:43] Infinity campaigns is the heart of what [14:46] campaign agents do. [14:48] The beauty of Infinity campaigns, [14:50] because it's built around the LLMs, is [14:52] that you can start personalizing [14:54] immediately. You don't have to spend [14:55] months building ML models. [14:58] You also don't need to use ML-based [14:59] decisioning, that takes a really long [15:01] time to start to work. [15:03] It provides exceptional campaign [15:05] performance out of the box, and of [15:07] course, batteries are included. We have [15:10] guardrails, you can do simulation runs [15:13] and measurement. Everything is built in. [15:17] And for the first time, [15:19] we're We're [15:20] a feature that so many of you have been [15:22] asking. [15:23] Reverse ETL, we can now have native [15:26] managed connectors [15:29] to multiple products in the advertising, [15:32] marketing, and customer experience [15:33] ecosystem [15:34] that with one click you can make it [15:36] work. [15:38] We're launching with approximately 20 [15:40] connectors to a rich ecosystem of [15:43] partners. We have dozen more connectors [15:45] coming by the end of the year [15:47] and many more beyond that. [15:51] Okay. [15:52] So, it's time to see Customer Lake in [15:55] action. So, I want to invite Justin to [15:57] give a demo of Customer Lake. Welcome, [16:01] Justin. [16:09] >> Thanks, Justin. [16:10] Oncore demo, I'm flattered. [16:12] Um [16:13] so, uh in this demo, um [16:16] we're going to give a quick overview of [16:17] campaign agents. [16:19] Um the assumption here is that you've [16:21] already used profile agents to ingest, [16:23] uh model, and resolve your customer [16:25] data. And what I'm going to walk through [16:26] is using campaign agents um to build [16:29] audiences using Genie and then Infinity [16:31] Campaigns to decision on a one-to-one [16:33] basis. Okay? So, let's jump into it. [16:37] Now, I'm a marketing manager for Brooks [16:39] Sport. And here I can see my campaigns [16:41] dashboard. This is my overview of my [16:43] active campaigns, some key audiences [16:45] that I'm tracking, um my most active [16:47] reverse ETL syncs, [16:50] um [16:51] kind of my kind of uh overview [16:53] dashboard. Let's jump over here, though, [16:56] into the audiences. [16:57] Here's a list of audiences that I've [16:59] created. I can see the size of these [17:00] audiences, the percent change over time. [17:03] Um what we're going to do is we're just [17:04] going to create a new audience. [17:06] And as I do this, I could create this [17:08] audience using uh rules and filters, but [17:11] we're going to use Genie. [17:13] And I'm going to copy and paste a prompt [17:15] in here. [17:16] Um include our most valuable Team USA [17:18] fans who have spent more than $500 on [17:20] merch this season, [17:21] um and exclude anyone who already bought [17:23] the USA fan kit, okay? So, let's see [17:25] what Genie comes up with. [17:27] Genie's going to go and actually [17:28] generate this audience, so it's going to [17:29] look at the attributes and the audiences [17:31] that I already have created. And there [17:33] we go, Genie generated this audience. [17:35] So, now I have an audience of 800,000 [17:37] customers. I can see the qualification [17:39] criteria here. If I wanted to, I could [17:41] go in here and tweak this, right? So, [17:43] this is [17:44] um uh editable. I'm going to trust Genie [17:47] and go ahead and accept this. [17:49] I come back out here, and now I have uh [17:52] some quick insights that Genie [17:53] generated. Um we'll come back to that in [17:55] a second. I actually want to show this. [17:57] Um you can view the SQL that is backing [18:00] this audience. So, this is starting to [18:02] look probably more familiar uh to many [18:04] of you in the audience. Uh if you've [18:05] ever supported marketing by pulling [18:07] lists and writing custom campaigns, um [18:10] you can get really complex SQL um to [18:12] power some of these campaigns, and we [18:14] got this from one prompt with Genie. [18:17] Um [18:19] let's go ahead and save this, um and [18:20] I'll get brought out to the main [18:22] insights dashboard here. [18:24] Um and I can see kind of insights on [18:26] this audience. These are automatically [18:27] generated by Genie. So, dynamically, [18:29] every audience is getting its own unique [18:31] dashboard. Um and then I can immediately [18:33] use this in a new campaign or even just [18:35] a new sync. Um so, the point is all [18:37] these audiences are actionable. This is [18:39] not just insights. You can immediately [18:41] run this in a campaign. Now, for the [18:42] sake of time, I'm going to go over here, [18:44] um and I'm going to find a live Infinity [18:48] campaign. Um in this example, it's my [18:51] in-venue fan experience. [18:53] Um so, let's quickly look into this [18:55] running campaign here, um and see how [18:57] it's configured. [18:59] All right. So, I can see the audiences [19:01] to include, the goals, [19:04] um conversion metrics, any sort of [19:05] guardrails that I wanted to apply. So, [19:07] this is kind of the constraints that [19:08] you're putting on the agent. I can see [19:11] the messages that I made available as [19:12] part of this Infinity campaign. So, here [19:14] I have a match day guide and a match day [19:17] offer and we'll come back to those in a [19:18] second. [19:19] Um [19:20] and the really interesting part is [19:21] looking at this agent's tab. [19:23] So, this is where the agent is actually [19:25] looking on a one-to-one basis um at each [19:28] individual customer. So, you can see a [19:29] list of sample profiles here. [19:31] Um here we have Marcus. Um this agent [19:35] decided to defer, so not send anything [19:36] to Marcus right now. Let's take a look [19:38] at why. We can actually see the [19:40] reasoning um behind this decision from [19:43] the agent. [19:44] What was evaluated? What was the timing [19:46] that was considered? Um what sort of [19:48] data signals were considered as part of [19:50] this decision? Um any sort of guardrails [19:52] that were applied, like I said, [19:53] frequency caps, etc. [19:55] Um if we [19:57] uh go into somebody who got the offer [20:00] here, we see this person got the match [20:03] day offer and same thing. Um [20:06] Genie decided to send this personalized [20:09] offer. And this is a preview of what [20:10] they actually got. So, this is a 15% um [20:14] offer off a seat upgrade. [20:16] Um let's find somebody who actually got [20:19] just the match day guide. So, here we [20:20] go. Um if we preview this, we can see [20:22] this customer got just a summary of the [20:25] match day kickoff, parking information, [20:27] etc. [20:29] So, campaign agents is personalizing [20:31] every single one of these message for [20:33] each individual customer in this [20:34] campaign continuously. [20:36] I can go out and see a overview of the [20:38] insights on this campaign. I can look at [20:41] performance by individual messages, [20:43] opens, clicks, conversions, etc. [20:46] Um to get kind of an aggregate view of [20:47] how this campaign is doing. [20:50] And with that, I'm back at the home [20:51] page. That's a quick run through of [20:53] Customer Lake Campaign Agents, [20:56] the world's first agentic CDP. [20:58] Thank you. Back to you, Tassa. [21:06] >> Okay. [21:08] Thank you, Justin. [21:10] So, the [21:12] uh the era of true one-to-one [21:14] personalization [21:15] is here. It's very, very, very, very [21:17] exciting. [21:18] So, in developing uh Customer Lake, we [21:22] were very fortunate and honored to work [21:26] with some of the top uh innovators [21:28] globally. [21:30] Um some of which uh were kind enough to [21:32] let us uh put their logos up on the [21:34] screen. These are world-class data and [21:36] marketing teams that are pushing the [21:39] boundaries of customer experience with [21:42] data, agents, and marketing strategy. [21:44] Thank you to all our early Customer Lake [21:47] partners. [21:48] And similarly, [21:50] um we have been uh building Customer [21:53] Lake surrounded by an elite ecosystem of [21:57] partners um that jointly serve our [21:59] customers in terms of data, content, [22:02] advertising, activation, and services. [22:06] Thank you to all of our partners that [22:07] are joining us today for this launch. [22:11] So, with that, I'm excited to say that [22:13] Customer Lake is in private preview [22:15] today. [22:16] Um if you're interested, please talk to [22:19] your account team. Thank you. [22:21] Uh please talk to your account team if [22:23] you want to learn more. We also have a [22:26] brand new CDP track in the breakouts. [22:31] And we also have a marketing forum today [22:34] at 1:00 p.m. You can find both Search [22:36] for Customer Lake in your uh Data and AI [22:40] Summit app, and you will see both the [22:43] marketing forum and the breakout [22:44] session. That's the place to learn more. [22:47] Some of the breakouts is from the team, [22:48] the engineers, the product folks that [22:50] actually built Customer Lake, and we [22:52] also have some of our early [22:54] uh customers talking about their first [22:56] experiences with Customer Lake as well. [22:59] Make sure to check it out. [23:02] So, with that, uh thank you for your [23:05] time. It's a very special moment for me [23:07] to be able to bring together the worlds [23:09] of business, customer experience, and [23:11] data all in one place [23:14] uh to build something that provides the [23:16] perfect customer experience a billion [23:18] times a day. [23:20] Thank you.